High Performance Scientific Computing complements research originally grounded exclusively on experimental investigations and analytical thinking. The power of efficient computational algorithms deployed on parallel high performance compute clusters enables exploration of complex phenomena at a virtually unrestricted space and time scales. In particular, computational models, calibrated to available observation data, allow robust predictions of future evolution of probable scenarios, supported by reliable quantification of the associated uncertainties and risks.
Focus areas of the group are the following:
- Parallel uncertainty quantification and propagation methodologies and numerical algorithms for deterministic and stochastic computational models
- Multi-level and multi-fidelity methods for optimal complexity in statistical sampling algorithms
- Massively parallel high performance computing, parallelization techniques and load balancing
- Numerical methods for hyperbolic nonlinear partial differential equations (shallow water, Euler, multi-phase)